4_32_rd
Browse files
README.md
CHANGED
|
@@ -15,8 +15,8 @@ license: apache-2.0
|
|
| 15 |
|
| 16 |
<a href='https://csgo-gen.github.io/'><img src='https://img.shields.io/badge/Project-Page-green'></a>
|
| 17 |
<a href='https://arxiv.org/abs/2404.02733'><img src='https://img.shields.io/badge/Technique-Report-red'></a>
|
| 18 |
-
[](https://huggingface.co/spaces/InstantX/
|
| 19 |
-
|
| 20 |
|
| 21 |
</div>
|
| 22 |
|
|
@@ -55,6 +55,15 @@ license: apache-2.0
|
|
| 55 |
This repo, named **CSGO**, contains the official PyTorch implementation of our paper [CSGO: Content-Style Composition in Text-to-Image Generation](https://arxiv.org/pdf/).
|
| 56 |
We are actively updating and improving this repository. If you find any bugs or have suggestions, welcome to raise issues or submit pull requests (PR) 💖.
|
| 57 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
## Pipeline 💻
|
| 59 |
<p align="center">
|
| 60 |
<img src="assets/image3_1.jpg">
|
|
@@ -143,12 +152,11 @@ controlnet_target_content_blocks = controlnet_BLOCKS['content']
|
|
| 143 |
controlnet_target_style_blocks = controlnet_BLOCKS['style']
|
| 144 |
|
| 145 |
csgo = CSGO(pipe, image_encoder_path, csgo_ckpt, device, num_content_tokens=4,num_style_tokens=32,
|
| 146 |
-
target_content_blocks=target_content_blocks, target_style_blocks=target_style_blocks,
|
| 147 |
controlnet_target_content_blocks=controlnet_target_content_blocks,
|
| 148 |
controlnet_target_style_blocks=controlnet_target_style_blocks,
|
| 149 |
content_model_resampler=True,
|
| 150 |
style_model_resampler=True,
|
| 151 |
-
load_controlnet=False,
|
| 152 |
|
| 153 |
)
|
| 154 |
|
|
|
|
| 15 |
|
| 16 |
<a href='https://csgo-gen.github.io/'><img src='https://img.shields.io/badge/Project-Page-green'></a>
|
| 17 |
<a href='https://arxiv.org/abs/2404.02733'><img src='https://img.shields.io/badge/Technique-Report-red'></a>
|
| 18 |
+
[](https://huggingface.co/spaces/InstantX/CSGO)
|
| 19 |
+
|
| 20 |
|
| 21 |
</div>
|
| 22 |
|
|
|
|
| 55 |
This repo, named **CSGO**, contains the official PyTorch implementation of our paper [CSGO: Content-Style Composition in Text-to-Image Generation](https://arxiv.org/pdf/).
|
| 56 |
We are actively updating and improving this repository. If you find any bugs or have suggestions, welcome to raise issues or submit pull requests (PR) 💖.
|
| 57 |
|
| 58 |
+
## Detail ✨
|
| 59 |
+
We currently release two model weights.
|
| 60 |
+
|
| 61 |
+
| Mode | content token | style token | Other |
|
| 62 |
+
|:------------:|:-----------:|:-----------:|:---------------:|
|
| 63 |
+
|csgo.bin|4|16| - |
|
| 64 |
+
|csgo_4_32.bin|4|32| Deepspeed zero2 |
|
| 65 |
+
|
| 66 |
+
|
| 67 |
## Pipeline 💻
|
| 68 |
<p align="center">
|
| 69 |
<img src="assets/image3_1.jpg">
|
|
|
|
| 152 |
controlnet_target_style_blocks = controlnet_BLOCKS['style']
|
| 153 |
|
| 154 |
csgo = CSGO(pipe, image_encoder_path, csgo_ckpt, device, num_content_tokens=4,num_style_tokens=32,
|
| 155 |
+
target_content_blocks=target_content_blocks, target_style_blocks=target_style_blocks,controlnet_adapter=True,
|
| 156 |
controlnet_target_content_blocks=controlnet_target_content_blocks,
|
| 157 |
controlnet_target_style_blocks=controlnet_target_style_blocks,
|
| 158 |
content_model_resampler=True,
|
| 159 |
style_model_resampler=True,
|
|
|
|
| 160 |
|
| 161 |
)
|
| 162 |
|